This project is just a demonstration of the neural networks algorithms, for real applications you should use a higher level framework or library.
In this project, I implemented a Neural Network using Python and NumPy. The parameters of the network are customizable in terms of the number and size of hidden layers, the size of the input and output layers, the parameters for learning and regularization.
The code implements:
- Forward Propagation
- Backpropagation
- Gradient Check (Very useful to test the implementation)
- Activation function (sigmoid)
To use the network you can simply instantiate an object from the NeuralNetwrok class with the required (or default) parameters, then supply the training data for training and then the testing data for prediction.
For an example on how to use the network, see "main.py" file for an implementation.